Three internal validation techniques are shown. Split-sample means that a random split is made, resulting in e.g. a 50% development and a 50% validation sample. Cross-validation uses the same principle, but alternates the development and validation samples (e.g. 50:50 split means 2 development and test rounds; a 90:10 split 10 rounds). The most extreme variant of cross-validation is the ‘jack-knife’, where n-1 patients are used for model development, with validation on the patient who was left out. The bootstrap is however the preferred technique (see next slide).